Eliary Inc. · Delaware C-Corp
One identity product, two surfaces. Lucid scans you with four frontier AIs and ships their disagreement as a share-card. Currot opens the questions only friends can answer, on the same persistent graph. AI starts understanding. Humans complete it.
Thesis · Signal Cost Theory
After Spence (1973): a signal is only meaningful if it's costly to fake. AI signals are free — they generate engagement but never relationships. As AI gets better, the cheap-signal supply explodes and the demand for costly, human signals rises with it.
Eliary builds for that gap. Lucid is the activation surface: four frontier AIs read you and disagree about who you are. Currot is the retention surface: the questions AI can't close, asked between people. Both surfaces share one persistent Curiosity Graph.
“The AIs say X. Your friends know Y.”
Lucid creates the engagement; Currot fulfills the relationship. The product is one job, two surfaces: figure out who I am with AI's help, then live that with people.
Why now
ask.fm crossed 100M+ users on a Q&A primitive before anonymous abuse burned the window shut. The user demand was real. The architecture was wrong.
AI is a neutral facilitator, not an anonymous abuser. Multi-AI scan absorbs the curiosity surface; humans get the parts that need humans. The wedge that was missing for a decade now exists.
Four frontier models reading the same person produce something inherently shareable: not consensus, but a debate about you. Card → friend tries → new card → share. The viral loop is the product, not bolted on.
Multi-AI personality scan. Four frontier models read the same person in parallel, and the disagreement is the share-card unit. The activation surface of the Eliary identity product — where engagement is created and the Curiosity Graph is seeded.
What 4 AIs think about you.
Answer the same 74 questions. Watch four frontier models classify you in parallel — sometimes converging, often clashing. The point isn't a verdict. It's the gap between four readings of one person.
After the scan, each AI becomes a conversation. Push back, follow up, ask the same question four ways. Generate a glassmorphic card with the agreement rate and four verdicts in one frame, and pass it on. The share is the loop.
Standalone-viable: same category as 16Personalities (100M+ users on identity tests), now with four real AIs and a social Portrait layer the legacy category could never build.
74 questions, four frontier models, one share card. The disagreement is the unit.
Card → friend tries → new card → share. The viral loop is the product, not bolted on.
65.7% of analyzed users score AI-skeptical (≤3 / 7, n=1,634). That's the real TAM, not the AI-easy segment.
Each AI becomes a separate thread. Push back, follow up, ask one question four ways.
Every scan writes to the Curiosity Graph. Surface 02 (Currot) reads from it. The retention is downstream.
How the two surfaces connect
Lucid and Currot are not two products. They are activation and retention on a single persistent identity graph. Architectural convergence: 90 days.
“Figure out who I am with AI's help, then live that with people.”
Currot is the human follow-through. Where Lucid surfaces what four AIs think, Currot opens the questions only friends can close — identity-first, in a calm space designed for relationship, not performance.
Ask Bold. Answer Real.
A social space built around one primitive: asking. Not posting, not broadcasting — asking. The question is the unit; the relationship is the outcome.
Send to one friend or many, choose who receives, and turn a question into an invitation that works even before your friend has the app. From close friends to communities to public figures, every conversation begins with a signed ask.
On the YC roadmap, the Ask primitive converges into Lucid as the Portrait-update retention layer in 90 days, sharing the same Curiosity Graph. Two surfaces, one persistent identity — 326 asks and 784 posts from real users, with 65% of cumulative signups active in the last 30 days.
This isn't traditional social media. It's the human side of the AI-mediated identity loop.
Questions, not posts. The relationship forms in the asking, not the broadcasting.
Friends don't need the app yet — the question itself is the invitation. Connection starts pre-signup.
Asks, answers, and Portrait updates persist across surfaces. Lucid scans, Currot fills in. Same graph.
Every ask is signed. AI is the mediator, not anonymity. Mutual-follow DMs, content-safety review, enhanced minor protection.
No algorithmic feed. No engagement scoring. The product has no incentive to manipulate attention.
Diagnostic Lab · 13-paper portfolio
We measure our own platforms with outside-researcher precision and publish the failures. Few founders can — fewer will. The diagnostic culture is the moat: it produces decisions before competitors notice the question.
Prospective case study of Currot's cold-start dynamics, measured in real time against a 10-CSV ledger. Documents what concentrated-creator distribution surfaces and what graph-native primitives the Curiosity Graph absorbs in v2.
Decomposes Lucid's mid-flow conversion across 4,041 sessions. Surfaces the 65.7% AI-skeptical finding (n=1,634, mean 2.62) — the resistant segment is the real TAM.
Formalizes engagement-signal devaluation under generative AI. Predicts the structural pressure that opens the human-signal market Currot occupies.
Preregistered RCT comparing human-only vs. AI-mixed relationship outcomes. OSF DOI pending. Tests the falsifiable form of Signal Cost Theory.
56 quarters of Meta SEC filings + 36 quarters of Snap + 15 question-platform lifecycles. Places both surfaces in lifecycle escape position. The architecture that survives builds the next surface before the current one ages out.
Founder · solo technical · self-taught
Solo founder of Eliary Inc. (Delaware C-Corp, incorporated July 3, 2025). Studied International Relations at SNU; learned engineering by shipping. Eighteen months of solo development, with two live products built in the final six weeks alongside a 13-paper theoretical series on platform senescence and signal cost. Every line by one person: Go backend on Spanner, Flutter for iOS/Android, Next.js web, four-provider AI orchestration.
The unusual move isn't the velocity — it's the rigor. The IR background is the signal-theory background. I measure my own platforms with outside-researcher precision and publish the failure modes. The diagnostic culture is the moat: it produces decisions before competitors notice the question.
“I am a founder who does the analysis other founders pay consultants for.”
Active co-founder search · #1 ask
Looking for a partner who has owned a consumer 0→1 growth loop end-to-end — the half of the business I have not done before. Engineering and product I can carry; mass-market consumer scaling I cannot.
Equity is irreversible. Slow-but-right is the discipline; the wrong partner permanently degrades the company. YC is the highest-quality channel for that search, which is the single biggest reason I'm applying to S26.
[email protected] →The bet
If Signal Cost Theory is right, Lucid + Currot compound on one Curiosity Graph and we own the AI-mediated identity category. If the unified thesis is wrong, Lucid still works alone as consumer multi-AI chat — a category 16Personalities serves at 100M+ users with no AI. Either branch is a real outcome. We are not pretending to product-market fit we don't have; we are documenting the path to it.